def create_model(self):
print("Creating model")
base_model = KerasInceptionV3(weights='imagenet', include_top=False, input_tensor=self.get_input_tensor())
# print("base_model.layers:", len(base_model.layers))
# self.make_net_layers_non_trainable(base_model)
x = base_model.output
x = GlobalAveragePooling2D()(x)
feature = Dense(config.noveltyDetectionLayerSize, activation='elu', name=config.noveltyDetectionLayerName)(x)
# x = Dropout(0.6)(feature)
predictions = Dense(len(Inference.classes), activation='softmax', name='predictions')(feature)
if config.isCenterLoss:
print(config.isCenterLoss)
input_target = Input(shape=(None,))
centers = Embedding(len(Inference.classes), 4096)(input_target)
print('center:', centers)
center_loss = Lambda(lambda x: K.sum(K.square(x[0] - x[1][:, 0]), 1, keepdims=True), name='center_loss')(
[feature, centers])
model = Model(inputs=[base_model.input, input_target], outputs=[predictions, center_loss])
elif config.isTripletLoss:
model = Model(input=base_model.input, output=[predictions, feature])
else:
print(base_model.input)
model = Model(input=base_model.input, output=predictions)
Inference.loaded_model = model
def _create(self):
base_model = KerasInceptionV3(weights='imagenet', include_top=False, input_tensor=self.get_input_tensor())
self.make_net_layers_non_trainable(base_model)
x = base_model.output
x = GlobalAveragePooling2D()(x)
x = Dense(1024, activation='elu', name=self.noveltyDetectionLayerName)(x)
predictions = Dense(len(config.classes), activation='softmax')(x)
self.model = Model(input=base_model.input, output=predictions)
def define(self, optimizer=Adam(lr=config.configured_learning_rate)):
self.optimizer = optimizer
keras_model = KerasInceptionV3(weights=None,
include_top=False,
input_tensor=self.get_input_tensor())
self.make_net_layers_non_trainable(keras_model)
#use standard model or fine turn model
self.fineturn(keras_model)